TRY_PARSE 但速度更快

Question

将大量（几乎 100 个）nvarchar 列转换为浮点数的最快方法是什么？

目前我有一堆 TRY_PARSE(column-name AS FLOAT) 语句，每列一个。

如果列不解析为浮动，我想要 NULL。没有错误通知，没有额外的智能。

有没有更快的方法？如果是的话，会是什么？

SQL Server 2012

Answer 1

TRY_PARSE只需一个简短的测试即可显示和之间的性能差异TRY_CAST

我的结果

• TRY_CAST327 毫秒
• TRY_PARSE17784 毫秒

这意味着，这TRY_CAST大约比TRY_PARSE（顺便说一句：性能TRY_CONVERT几乎相同）快 50 倍。TRY_CAST

尝试一下并发布你的结果;-)

SET LANGUAGE ENGLISH;

DECLARE @tbl TABLE(SomeFloatAsString VARCHAR(100),fl2 VARCHAR(100),fl3 VARCHAR(100),fl4 VARCHAR(100),fl5 VARCHAR(100),fl6 VARCHAR(100),fl7 VARCHAR(100),fl8 VARCHAR(100),fl9 VARCHAR(100),fl10 VARCHAR(100));
DECLARE @i INT=0;
WHILE @i<100000
BEGIN
    INSERT INTO @tbl VALUES(RAND(),RAND(),RAND(),RAND(),RAND(),RAND(),RAND(),RAND(),RAND(),RAND());
    SET @i+=1;
END

DECLARE @d DATETIME2 = SYSUTCDATETIME();

SELECT TRY_CAST(SomeFloatAsString AS FLOAT) fl1
      ,TRY_CAST(fl2 AS FLOAT) fl2
      ,TRY_CAST(fl3 AS FLOAT) fl3
      ,TRY_CAST(fl4 AS FLOAT) fl4
      ,TRY_CAST(fl5 AS FLOAT) fl5
      ,TRY_CAST(fl6 AS FLOAT) fl6
      ,TRY_CAST(fl7 AS FLOAT) fl7
      ,TRY_CAST(fl8 AS FLOAT) fl8
      ,TRY_CAST(fl9 AS FLOAT) fl9
      ,TRY_CAST(fl10 AS FLOAT)fl10
INTO #tmp1
FROM @tbl

SELECT DATEDIFF(MILLISECOND,@d,SYSUTCDATETIME())


DECLARE @d2 DATETIME2 = SYSUTCDATETIME();

SELECT TRY_PARSE(SomeFloatAsString AS FLOAT) fl1
      ,TRY_PARSE(fl2 AS FLOAT) fl2 
      ,TRY_PARSE(fl3 AS FLOAT) fl3
      ,TRY_PARSE(fl4 AS FLOAT) fl4
      ,TRY_PARSE(fl5 AS FLOAT) fl5
      ,TRY_PARSE(fl6 AS FLOAT) fl6
      ,TRY_PARSE(fl7 AS FLOAT) fl7
      ,TRY_PARSE(fl8 AS FLOAT) fl8
      ,TRY_PARSE(fl9 AS FLOAT) fl9
      ,TRY_PARSE(fl10 AS FLOAT)fl10
INTO #tmp2
FROM @tbl

SELECT DATEDIFF(MILLISECOND,@d2,SYSUTCDATETIME());